How to benchmark hardware in Linux vs Windows without software limits?

Utini · 2019-05-27 07:23:47

Hello everyone,

my laptop is setup with Archlinux and Windows 10. Both OS are (to me) completely setup and optimized for my usage.
How ever, in video rendering I seem to notice performance differences between the two OS.
This could be related to software, codec, drivers, .... but I want to make sure that it e.g. is not related to TLP, selected scheduler,...
So I thought it would be interesting to benchmark both OS in a way that will not be compromised by different drivers or codecs. Something that stresses CPU, NVME SSD, RAM and maybe the iGPU (but preferable not dGPU) but is not limited to different architecture and programming limits. E.g. 7-zip in Windows utilizes all 12 CPUs while on Linux most archivers only use 1 CPU. FFMPEG seems to be fast on Windows than on Linux, no idea why, but it is software and not hardware related and there for bad as benchmark.

Can anyone recommend me on how to do this kind of benchmark?

Last edited by Utini (2019-05-27 08:16:26)

d_fajardo · 2019-05-27 08:08:06

You can have a look at phoronix test suite.

Utini · 2019-05-27 08:11:23

I already looked and phoronix benchmarks but it uses exactly the software that I mentioned which simple work different depending on the OS.
E.g. ffmpeg and 7-zip.

It's code/features/dependencies are too different between Windows/Linux which causes slower results eventhough the hardware is the same.
So for some benchmarks it mostly is benchmarking the software instead of the hardware.

seth · 2019-05-27 14:30:23

To rule out the SW you'll have to use the same SW and that SW must use the same code paths. You can try some windows benchmark via wine (ack'ing a small discount for the wine mappings)

As for 7z : "7z --help", look out for the -mmt option…

Edit: one dash only…

Last edited by seth (2019-05-27 14:30:54)

Lone_Wolf · 2019-05-28 11:02:01

Maybe the tests used in https://www.phoronix.com/scan.php?page= … readripper would be suitable ?

Utini · 2019-05-29 06:45:28

Lone_Wolf wrote:

Maybe the tests used in https://www.phoronix.com/scan.php?page= … readripper would be suitable ?

Thanks but I want to run the benchmark on my system to test my existing configuration.
E.g. with TLP, scheduler, etc etc are all set correctly and give me the performance that I would also get on Windows (which I have also setup "correctly").

DrZaius · 2019-05-31 19:33:21

Utini wrote:

FFMPEG seems to be fast on Windows than on Linux, no idea why, but it is software and not hardware related and there for bad as benchmark.

Show your command and the complete console output. Use the same input and command on each OS. Also use the same ffmpeg version. If encoding with an external library via ffmpeg, such as libx264, then also use the same library version.

Arch Linux

#1 2019-05-27 07:23:47

How to benchmark hardware in Linux vs Windows without software limits?

#2 2019-05-27 08:08:06

Re: How to benchmark hardware in Linux vs Windows without software limits?

#3 2019-05-27 08:11:23

Re: How to benchmark hardware in Linux vs Windows without software limits?

#4 2019-05-27 14:30:23

Re: How to benchmark hardware in Linux vs Windows without software limits?

#5 2019-05-28 11:02:01

Re: How to benchmark hardware in Linux vs Windows without software limits?

#6 2019-05-29 06:45:28

Re: How to benchmark hardware in Linux vs Windows without software limits?

#7 2019-05-31 19:33:21

Re: How to benchmark hardware in Linux vs Windows without software limits?

Board footer